Playstation 5 [PS5] [Release November 12 2020]

[Arwin]

The internets is fickle and changes its mind all the time.

PS4: Performance is everything

PS5: Performance doesn't matter

- the internets

For me personally, you could have had the same discussion with the PS4 Pro. It’s why backward compatibility is important.

That said, this time the performance isn’t all that easy to judge. The previous generation were basically budget PCs with GPU and bandwidth being the main differentiating factors. This time, all aspects of performance have been much more carefully tuned with the highlight being an incredibly huge difference in solid state innovation - that is one of the biggest jumps in performance on an important aspect of the console design ... ever? And in that area the difference is much bigger percentage wise than differences in GPU or bandwidth. So it is much harder to judge the actual overall performance differences against previous generations. I think it is probably no surprise that one is focusing more on cross gen than the other though.

Incidentally, if the PS5 is going to support slotting in an additional NVME drive completely inside its form factor, then that will explain part of its size right there, assuming this is potentially also an additional part that needs cooling.

 

[3dilettante]

No, it won't happen. The variable frequency system is based on number of instructions usage, not power usage. The variable clocks are actually not implemented using a power cap.

AMD's hardware uses a mixture of activity counters and blocks of logic that perform representative operations of the larger logic blocks as inputs into a function that estimates power consumption. It may not be measuring the cycle-level consumption directly, but there's still a power target or max.

Developers will fix the activity of GPU and CPU in a way that is sustainable... not the max possible... it's a way to unload to developer the responsibility of making games that doesn't break the console in 50 Celsius in India ? Or maybe 40 Celsius in Italy ???

If you look at the specs for prior consoles, there's an operating temperature range. The PS4's is 5-35C.
At the temps you give, Sony promises nothing. The PS5's size and cooling may be provisioned for the upper range of the operating temperature. Developers rely on the PS5 to control its own consumption. Any profiling they do is for determining if the game performs as expected, which Sony's specs say is temp-independent.

 

[Xbat]

Developers will fix the activity of GPU and CPU in a way that is sustainable... not the max possible... it's a way to unload to developer the responsibility of making games that doesn't break the console in 50 Celsius in India ? Or maybe 40 Celsius in Italy ???

No, the consoles fans will spin up more or the machine will shutdown but performance will stay the same.

 

[HBRU]

Developers won't sell games that make the console too loud ? Or worst that let it stop for protection ? Or maybe they will push further as slim comes to the market... ?? I love many aspects of day one PS5 but this may be possibly IMHO a problem. The max frequency is very high....

 

[Shifty Geezer]

Developers will fix the activity of GPU and CPU in a way that is sustainable... not the max possible... it's a way to unload to developer the responsibility of making games that doesn't break the console in 50 Celsius in India ? Or maybe 40 Celsius in Italy ???

That goes against what Cerny said in the post you are replying to!

"It wouldn't do to run a console slower simply because it was in a hot room"

Devs have never cared about tweaking their games to fit in environmental conditions the consoles isn't made for before (they don't even optimise their title screens/menus to stop the fans whirring up!), and they aren't going to start now. Just like now, when the consoles gets hot, the fan will rev up. The console will not slow down because it's hot and, exactly as Cerny expressly stated so people in hot countries wouldn't be worried, everyone will get the same performance from their PS5.

Developers won't sell games that make the console too loud ? Or worst that let it stop for protection ? Or maybe they will push further as slim comes to the market... ?? I love many aspects of day one PS5 but this may be possibly IMHO a problem. The max frequency is very high....

No, no and no. It's no different to current consoles. Regardless how the power is moved between CPU and GPU, it won't throttle down and devs are free to push the hardware as much as they care to, to the point perhaps that the fan will sound like a jet engine in menus once again in hot environments.

 

[London Geezer]

Developers won't sell games that make the console too loud ? Or worst that let it stop for protection ? Or maybe they will push further as slim comes to the market... ?? I love many aspects of day one PS5 but this may be possibly IMHO a problem. The max frequency is very high....

Again. Power consumption is CONSTANT (while frequencies will fluctuate). So the PS5 will never over-heat like current consoles do, requiring extra cooling and resulting in noise. Unless it's broken.

So this will never, ever happen.

This has been explained so many times here it's getting a bit annoying having to reiterate this over and over again.

 

[HBRU]

If Sony says that at max 35 celsius you as developer can use the max GPU+CPU power... maybe as developer want to extend trouble free that max temperature by using a bit less the GPU+CPU... this is what is going to happen IMHO.

 

[Shifty Geezer]

If Sony says that at max 35 celsius you as developer can use the max GPU+CPU power... maybe as developer want to extend trouble free that max temperature by using a bit less the GPU+CPU... this is what is going to happen IMHO.

You're free to believe that despite everything said to the contrary, but that's the end of this discussion now to stop it becoming cyclic and return the discussion to other matters.

 

[iroboto]

Again. Power consumption is CONSTANT (while frequencies will fluctuate). So the PS5 will never over-heat like current consoles do, requiring extra cooling and resulting in noise. Unless it's broken.

So this will never, ever happen.

This has been explained so many times here it's getting a bit annoying having to reiterate this over and over again.

Uhhh. It can still overheat. PS5 is boost mode that is based on fixed power draw and activity level. They design a thermal system around the fixed power draw and worst case activity level, but that doesn’t make it impossible to overheat. It doesn't mean fans won’t get louder; they must work harder as temperatures keep going up.

Remember, ps5 won’t down-clock as a result of heat, it downclocks as a result of activity
Level. Which means someone playing in the desert vs someone playing in the North Pole; their consoles will perform the same, but the fan levels will be different.


having said that, it shouldn’t overheat. And neither should XSX in the same vein. But clogged consoles and bad environments, poor manufacturing makes for any possibility.

 

[zupallinere]

My analogy. The PS5 is a tall quick sprinter while the XBSX is a short powerful power lifter. Both are cool.

Both designs have solid functional elements that drive their look which is something that I appreciate. The Series X looks like an air duct (as opposed to those who see a duck in the PS5) and owns it's stout proportions. The PS5 almost looks like a 3D visualization of the air going through it.
It looks like a router of course but it reminds me a bit of the Spindrift which was the ship in the old SF TV series Land of the Giants

or a baleen whale maybe.

 

[Shifty Geezer]

Uhhh. It can still overheat....having said that, it shouldn’t overheat.

For clarity, as I don't want to remove your post for violating the End-Of-Discussion as it's a general remark on console design, the console can still get hotter, causing the fans to spin up to cool it down, and, worst case, the console may, in extreme circumstances, overheat and power down, like every console.

And neither should XSX in the same vein.

Indeed, unless a console incorporates thermal throttling and reduces the processor speeds to keep temperatures down, giving people in hot environments a worse experience, all consoles will operate the same in all temperatures so long as the cooling solution can work sufficiently within that ambient temperature, revvnig the fans up to crazy speeds if needed. And when the cooling cannot work, the console will power down - it won't play at reduced framerates.

Each console will have different cooling capacity and associated fan noise, which is as yet unknown for the upcoming machines. There is nothing different in PS5 to every other consoles in how it manages system heat. It does not feature thermal throttling.

 

[iroboto]

For clarity, as I don't want to remove your post for violating the End-Of-Discussion as it's a general remark on console design, the console can still get hotter, causing the fans to spin up to cool it down, and, worst case, the console may, in extreme circumstances, overheat and power down, like every console.
Indeed, unless a console incorporates thermal throttling and reduces the processor speeds to keep temperatures down, giving people in hot environments a worse experience, all consoles will operate the same in all temperatures so long as the cooling solution can work sufficiently within that ambient temperature, revvnig the fans up to crazy speeds if needed. And when the cooling cannot work, the console will power down - it won't play at reduced framerates.

Each console will have different cooling capacity and associated fan noise, which is as yet unknown for the upcoming machines. There is nothing different in PS5 to every other consoles in how it manages system heat. It does not feature thermal throttling.

Yup. Indeed neither console does thermal throttling; so they will both perform until
Shutdown, which was what I wanted to say.

Sorry, I didn’t realize you guys were talking to someone I put on my ignore list. Quotes don’t appear, I had to double back to see what actually happened.

 

[dobwal]

No, it won't happen. The variable frequency system is based on number of instructions usage, not power usage. The variable clocks are actually not implemented using a power cap.

I miss that piece of info where did you get that from?

In Road to the PS5, Cerny explicitly states that instead of constant frequency with varying power consumption based on workload, the PS5 basically runs at constant power with varying frequency based on workload. The cooling design is designed for that constant power level.

 

[eastmen]

I have a hard time believing that the PS5 is going to have any issue with noise.

Prior to the XB1, MS had similar issues with the 360. They readily addressed the issue the very next gen. Given how Sony capped the power consumption of the SOC and was willing to accept a rather large form factor, I can’t see Sony stumbling because of poor quality control of its suppliers. It would be a massive waste of effort.

Noise issues have been a problem since the ps3 generation. Over time the consoles would become much more noisy due to either games making better use of the hardware and it running hotter or dust build up or both. So I am hopefully sony has addressed the issue

 

[Globalisateur]

I miss that piece of info where did you get that from?

In Road to the PS5, Cerny explicitly states that instead of constant frequency with varying power consumption based on workload, the PS5 basically runs at constant power with varying frequency based on workload. The cooling design is designed for that constant power level.

You just wrote a great summary !
"varying frequency based on workload" [meaning some max number of instructions, the workload, will ensure some max power is theoretically never exceeded]
"cooling designed for constant power" [designed in the factory for a theoretical max power]

 

[dobwal]

You just wrote a great summary !
"varying frequency based on workload" [meaning some max number of instructions, the workload, will ensure some max power is theoretically never exceeded]
"cooling designed for constant power" [designed in the factory for a theoretical max power]

And so you think the PS5 constantly profiles different instructions and the volume of those instructions in flight across the SOC’s cpu/gpu/io devices and continuously sets frequency to limit power consumption?

Why go through all that when you can simply set a power cap? Basically a limit of how much power the SOC can pull. What’s the benefit to do it based on instruction count?

 

[London Geezer]

Uhhh. It can still overheat. PS5 is boost mode that is based on fixed power draw and activity level. They design a thermal system around the fixed power draw and worst case activity level, but that doesn’t make it impossible to overheat. It doesn't mean fans won’t get louder; they must work harder as temperatures keep going up.

Remember, ps5 won’t down-clock as a result of heat, it downclocks as a result of activity
Level. Which means someone playing in the desert vs someone playing in the North Pole; their consoles will perform the same, but the fan levels will be different.


having said that, it shouldn’t overheat. And neither should XSX in the same vein. But clogged consoles and bad environments, poor manufacturing makes for any possibility.

In my defence, I did write that it won’t overheat like current consoles do. As in, unless it’s faulty or clogged or whatever, it won’t overheat in the same situations we experience now - for example in menus or other very random situations - which is what I think you agree with.

Anyway, I think Shifty has declared this a redundant discussion, and also I don’t really care so *shrugs and hugs*

 

[3dilettante]

And so you think the PS5 constantly profiles different instructions and the volume of those instructions in flight across the SOC’s cpu/gpu/io devices and continuously sets frequency to limit power consumption?

Why go through all that when you can simply set a power cap? Basically a limit of how much power the SOC can pull. What’s the benefit to do it based on instruction count?

While there are likely many contributors to the power management system, a major component is that the physical and electrical behavior of the silicon is profiled at various points in the chip's operating range, to give its thermal response at various temperatures, voltages, and clocks. This profile data is looked up when a given activity counter registers an action, to get an estimate of the power consumption that resulted from the operation. Some later additions to the method include dummy ALUs or partial register file blocks that periodically perform some kind representative activity to get a tighter approximation of what the hardware is consuming. This isn't the same as instruction count, since instructions can translate into different levels of activity or different internal operations. An instruction may or may not generate multiple cache misses, or might wake up blocks that were clock-gated or worse power-gated, and all those have varying power costs.

The original motivation for this stems from the question of measuring temperature or power consumption for the silicon at small time scales and in any hot spots. The chip has thermal limits to safe operation, which traditionally required a wide safety margin for worst-case scenarios. Using temperature in a single thermal sensor doesn't cover the whole chip, and thermal sensors are relatively large to put them by every active block. Their response time and the speed that heat travels from a hot spot to a sensor can be a problem since local power spikes can push local temperatures from nominal to dangerous in millisecond or shorter time frames.
AMD's approach was to profile the silicon for how it reacted to events at different places in the performance envelope, and then used the activity count to generate an estimate. While conservative, it was based on a dynamic approximation that worked at microsecond ranges, rather than an estimate decided at product design time with very wide safety margins.
Since then, AMD may have also done more local electrical monitoring as well, which can produce more accurate estimates of power consumption and can tighten the estimate of how much a given region can heat up based on how much additional power can be consumed in a given time step.

The latest boost functionality and high clocks for Zen come from pushing silicon to near the safe limits of voltage and temperature of the silicon process for controlled periods of time, which under prior methods could very quickly overrun them. Sony's method appears to piggy-back on a lot of this work, and its operating point discards much of that boost range. The physical characterization tables seem to be tuned for consistency, which means the power management hardware thinks the silicon has a given baseline set of properties, regardless of whether the silicon itself can do better.
It's possible that AMD's method is over-engineered for what Sony is doing, but it would be more work to take it out at this point.

 

[iroboto]

In my defence, I did write that it won’t overheat like current consoles do. As in, unless it’s faulty or clogged or whatever, it won’t overheat in the same situations we experience now - for example in menus or other very random situations - which is what I think you agree with.

Anyway, I think Shifty has declared this a redundant discussion, and also I don’t really care so *shrugs and hugs*

oh yea.
i'd be surprised if today's consoles overheated in that type of situations (menus and maps, etc)

Usually shouldn't overheat, that would have to be a really poor console. Fans may spin loud though. But there's no way to guarantee relief from this type of thing as we're trying to get the most performance with the cheapest amount of cooling.

 

[iroboto]

And so you think the PS5 constantly profiles different instructions and the volume of those instructions in flight across the SOC’s cpu/gpu/io devices and continuously sets frequency to limit power consumption?

Why go through all that when you can simply set a power cap? Basically a limit of how much power the SOC can pull. What’s the benefit to do it based on instruction count?

activity level and instructions are not the same. What causes increased power is when bits flip values. 0 -> 1 and 1s -> 0s.
The more registers and ALU that is flipping throughout the chip, the more power that is drawn.

They did set a power cap. But they wanted to have boost as well. So they based the boost off the activity level, not the thermal level.